Don’t forget to upload the logs.
FWIW to get face recovery to work under Win10 on this image I had to switch off Recover Faces then back on. Without that only the remove noise worked.
Seems like it was released a bit too fast, maybe we get the fix today.
That’s a good idea. I tried switching it on/off too. Unfortunately, it didn’t fix the problem for me. I also tried turning on/off noise removal and sharpen. I tried changing CPU/GPU settings too.
I don’t think I mentioned which version of Monterey I’m running, it’s 12.6.5.
Found a solution for my Windows 10 Sytem.
It was a “accidential coincidence” (in my case).
Problem: after update to 1.3.4 photoai crashed during saving procedure. I’d tried several ways to downgrade to elder versions. Even 1.2.3. But every version crashed now during savin’ procedure. I made a complete clean install. Cleaned up harddisk and registry. No effekt. Same issue.
Yesterday: I looked for the right RESTORE POINT in Windows.
I found the one right after updating to 1.3.4. Windows showed a MESSAGE: FOLLOWING APP IS FAULTY AND NEED TO BE REINSTALLED. (TopazPhotoAI). I tried it. It seems that Windows made a new clean install of verison 1.3.4. Everything is working now like with version 1.3.3 in version 1.3.4. BUT I dont want to acivate the new beta functions. I left it basicaly look&feel like 1.3.3
Hope that can help at least some WINusers
After restarting PAI on my system the face recovery would not work again (I usually hibernate the system so PAI is rarely restarted). No amount of switching on/off face recovery would make it work. Changing the enhance resolution from 4x to 2x made it work this time. There appears to be some issue with the size of the PAI window, resolution and face recovery. With higher resolution settings and a smaller PAI window the preview is not getting updated. Changing the preview size to less than 100% also helps.
My system is Win10, Xeon W2123, 32GB with Quadro P2000 at 2560x1440.
Interesting YT video on update 1.3.4 new (beta) remove noise models: TOPAZ PHOTO AI (New Update Version 1.3.4) Now with 2 NEW (REMOVE NOISE) Beta Models - YouTube
I tried your image on my Macmini M1 w/latest Ventura and it initially didn’t work… I switched off Subject detection and it worked… then I switched Subject detection back ON and it continued to work… at that point, regardless of settings, it continued to work… there appears to be something hinky in the way it refreshes…
I backed out without saving anything and started over again, and now it won’t enhance the two faces regardless of what is turned on or off…
In fact, now, after testing… Face Recovery is borked…
I turned everything off, and Face recovery is actually degrading the Face…
This is on an M1 Mac Mini w/Latest Ventura
Additionally, this occurs in both Stand-alone mode, and from the LR plug-in Edit In mode…
Well, I installed the latest version anyway. The good thing is that I can save DNG files again. The negative point is that I don’t have the new denoise Beta models on the DNG. See screenshot and that the quality is not great with the current models.
Also, the colors are not those of the thumbnail or the Tiff generated by Lightroom if I go into Photo AI without retouching the image.
the 2 images are in DNG. The one on the left is the original and the other one is the output in preserved format
Another black point, the DNG photos of type HDR always have this defect of stretching on the edges.
See screenshot on the right
On a jpeg, I can see the new denoise models.
On the Raw ORF files I don’t have access to the new denoise models either. So, it looks like it’s only on non-Raw files. On Tiffs, it works.
Conclusion, since Lightroom, it is always preferable to make open by sending as a tiff file in order to have more possibility. I hope that in the future versions of the software all its inconveniences will be corrected.
I just noticed that the face recovey tool is much less efficient than the previous version. In 90% of the cases, between the tool activated or not, there is no difference whereas Gigapixel recovers the quality of the faces really well.
Big example here
Photo AI 1.3.4
Gigapixel 6.3.3
1 Like
I gave the on/off subject detection a try, but it didn’t get the face recovery to work for me. I tried selecting one face or the other too, instead of both at once. I’ve gone through turning on/off other features too without finding a solution.
There’s definitely a fly in the code ointment somewhere. Even for those that got it working on their set-ups, you shouldn’t have to turn features on/off randomly to coax the software into working.
This is frustrating and I hope they could roll out a fix.
Yazi is working on it…
Fails to write to the directory
i verified the directory exists and is writeable yet the install program continues to fail.
1.3.4 (also 1.3.3) is creating a vignette on RAW images from a Canon G7X MKIII. Not sure where or how to upload images.
This is being fixed. Please update to v1.3.5 which should be released on this Thursday and this issue should be fixed.
1 Like
After Running some RAW files Photo AI no longer functions. Good thing I have Denoise also a reliable solution!
Something to do with Controlled Folder Access? Search for Controlled Folder Access and look at the Block history. You should be able to enable access if it is that. Controlled Folder Access first appeared in Win10 but was not enabled by default.
Bang on - I had this same issue a while back. Just add the program to the controlled folder access.
Hello! I can partially confirm this behavior.
1.3.4 crashes as soon as Denoising is set to strong (old) (when I select “Auto” as AI processor). It’s probably using my NVIDIA GeForce GTX 1660 GPU here. As soon as I set to “CPU” the crash doesn’t happen.
2023-05-16-14-56-16.tzlog (661.8 KB)
However, the strong Denoising is not satisfactory either, even if I set the strength to 100% (see image):
The area around the face is well denoised, the rest of the picture hardly at all.
Denoising v2 doesn’t offer any real improvement either. Due to the formation of the filaments in the dark area on the right, the quality seems even worse to me:
1 Like
